In this section, I will quantify and visualize group differences and longitudinal changes in the relative abundance of targeted taxa.
I will start with a high level look at general taxonomic groups, and then zero in on taxa in the literature that have previously been identified as differentially abundant either in murine pregnancy or in murine models of metabolic syndrome or GDM.
I will then do some visualizations of overall taxonomic composition
Changes / differences in individual taxa can be presented a number of ways. Below shows some experimentation with different visualization styles — I’m undecided on which tell the “story of the data” best.
Additionally, due to my caution around my Day 8 samples, I will sometimes opt to visualize with D5 set as endpoint.
First let’s investigate broad strokes changes in the relative abundances of two major bacterial phyla: Bacteroidetes and Firmicutes.
For our first plot, I’ll visualize relative abundance in each group
across all days
We can also show each hormone in it’s own panel
Next I’ll visualize this as baseline to intervention changes
I’m curious to visualize as a bar graph showing changes from baseline (D-3) to endpoint (D8).
To be comprehensive, I will then do the same with baseline (D-3) to
endpoint (D5), since the D8s are suspicious.
Stats testing is hidden from this html output, but these are the key
results:
No paired t-tests are significant for differences in Bacteroidetes relative abundance at D-3 and D8 for any of the hormone groups
For differences in Bacteroidetes relative abundance at D-3 and D5, the paired t-test is significant for PL mice (p = 0.036) but not for saline or PGH.
Next I’m going to calculate and plot the absolute change in Bacteroidetes % for each mouse
Change in Bacteroidetes from D-3 to D8 (i.e., D8 rel abundance - D-3
rel abundance)
The stats analyses are hidden in this HTML, but with this group comparision in Bacteroidetes change, error bars are high enough that no two groups reach significance in difference.
Next I’m going to repeat everything that I did above, but for Firmicutes.
Stats testing is hidden from this html output, but these are the key
results:
No paired t-tests are significant for differences in Firmicutes relative abundance at D-3 and D8 for any of the hormone groups
For differences in Firmicutes relative abundance at D-3 and D5, the paired t-test is significant for PL mice (p = 0.035) but not for saline or PGH.
This matches the results found for Bacteroidetes.
Though its a bit of an outdated metric, we can also quantify the B:F
ratio:
Those have some massive error bars on the baseline days!
Stats testing is hidden from this html output, but these are the key results:
No paired t-tests are significant for differences in Bacteroidetes relative abundance at D-3 and D8 for any of the hormone groups
No paired t-tests are significant for differences in Bacteroidetes relative abundance at D-3 and D5 for any of the hormone groups, but for PL mice, it is approaching significance (p = 0.098)
To increase my sample size, I’m going to combine D-3 and D0 into
baseline and have D3 and beyond as intervention:
Stats testing is hidden from this html output, but interestingly, it seems that because the baseline error bars are so high, there is no significant difference between baseline and intervention for F:B ratio for any of the hormone groups. Potentially also because these had to be done as unpaired rather than as paired t-tests, reducing our power.
I decided to look at the following taxonomic groups based on what I’d
seen in the literature:
I’m not going to show every possible plot for each of the above 7 taxa, but I’lll show some highlights for each below: (complete rel and abs abundance analyses are found in Grace’s HEB 114 Lab 2 Draft RMarkdown
This might look like a bloom in PL mice, but when we scrutinize more
closely, it’s likely driven by a single individual (one in PL and one in
saline), rather than being a consistent trend:
From this plot, I decided that when abundances are predominantly low,
it’s better to plot each mouse individually to see if blooms are
constrained to one individual.
Seems to bloom in all groups on day 8 (day with likely
contamination):
Given the distrust of D8 samples, I’m going to filter them out:
(Note: There was one clear crazy outlier mouse, so I filtered it
out)
This also increased across groups on the day of likely contamination
But if you look at relative and absolute abundace without D8, it
might be more interesting:
It looks like it might decrease in PGH? But unclear if this is
significant at all…
This is another that spikes in absolute abundance on D8 in all
groups:
Below I’ll show relative abundance both with and without the outlier
mouse (just by changing y-axis scale)
So this taxa increases in several of the PL individuals (3), although
one mouse started with an abnormally high concentration?
Seems to have decreased in PL mice, but started from baseline at an
abnormally high level.
I’m going to start by looking at phylum-level taxonomic composition across groups.
Because we saw potential shifts in the F:B ratio of PL mice, let’s start at the phylum level:
I’ll now do the same for the levels of class, order, family, and
genus: